Modelling competing risks data with missing cause of failure

Giorgos Bakoyannis, Fotios Siannis, Giota Touloumi

Research output: Contribution to journalArticle

19 Citations (Scopus)

Abstract

When competing risks data arise, information on the actual cause of failure for some subjects might be missing. Therefore, a cause-specific proportional hazards model together with multiple imputation (MI) methods have been used to analyze such data. Modelling the cumulative incidence function is also of interest, and thus we investigate the proportional subdistribution hazards model (Fine and Gray model) together with MI methods as a modelling approach for competing risks data with missing cause of failure. Possible strategies for analyzing such data include the complete case analysis as well as an analysis where the missing causes are classified as an additional failure type. These approaches, however, may produce misleading results in clinical settings. In the present work we investigate the bias of the parameter estimates when fitting the Fine and Gray model in the above modelling approaches. We also apply the MI method and evaluate its comparative performance under various missing data scenarios. Results from simulation experiments showed that there is substantial bias in the estimates when fitting the Fine and Gray model with naive techniques for missing data, under missing at random cause of failure. Compared to those techniques the MI-based method gave estimates with much smaller biases and coverage probabilities of 95 per cent confidence intervals closer to the nominal level. All three methods were also applied on real data modelling time to AIDS or non-AIDS cause of death in HIV-1 infected individuals.

Original languageEnglish (US)
Pages (from-to)3172-3185
Number of pages14
JournalStatistics in Medicine
Volume29
Issue number30
DOIs
StatePublished - Dec 30 2010
Externally publishedYes

Fingerprint

Competing Risks
Multiple Imputation
Grey Model
Modeling
Proportional Hazards Model
Missing Data
Proportional Hazards Models
Cause-specific Hazard
Cumulative Incidence Function
Estimate
Missing at Random
Data Modeling
Coverage Probability
Simulation Experiment
Categorical or nominal
Confidence interval
HIV-1
Cause of Death
Acquired Immunodeficiency Syndrome
Confidence Intervals

Keywords

  • Competing risks
  • Cumulative incidence
  • Fine and Gray model
  • Missing cause of failure
  • Multiple imputations

ASJC Scopus subject areas

  • Epidemiology
  • Statistics and Probability

Cite this

Modelling competing risks data with missing cause of failure. / Bakoyannis, Giorgos; Siannis, Fotios; Touloumi, Giota.

In: Statistics in Medicine, Vol. 29, No. 30, 30.12.2010, p. 3172-3185.

Research output: Contribution to journalArticle

Bakoyannis, Giorgos ; Siannis, Fotios ; Touloumi, Giota. / Modelling competing risks data with missing cause of failure. In: Statistics in Medicine. 2010 ; Vol. 29, No. 30. pp. 3172-3185.
@article{b1183e950a724a4c92b83493fe4013ac,
title = "Modelling competing risks data with missing cause of failure",
abstract = "When competing risks data arise, information on the actual cause of failure for some subjects might be missing. Therefore, a cause-specific proportional hazards model together with multiple imputation (MI) methods have been used to analyze such data. Modelling the cumulative incidence function is also of interest, and thus we investigate the proportional subdistribution hazards model (Fine and Gray model) together with MI methods as a modelling approach for competing risks data with missing cause of failure. Possible strategies for analyzing such data include the complete case analysis as well as an analysis where the missing causes are classified as an additional failure type. These approaches, however, may produce misleading results in clinical settings. In the present work we investigate the bias of the parameter estimates when fitting the Fine and Gray model in the above modelling approaches. We also apply the MI method and evaluate its comparative performance under various missing data scenarios. Results from simulation experiments showed that there is substantial bias in the estimates when fitting the Fine and Gray model with naive techniques for missing data, under missing at random cause of failure. Compared to those techniques the MI-based method gave estimates with much smaller biases and coverage probabilities of 95 per cent confidence intervals closer to the nominal level. All three methods were also applied on real data modelling time to AIDS or non-AIDS cause of death in HIV-1 infected individuals.",
keywords = "Competing risks, Cumulative incidence, Fine and Gray model, Missing cause of failure, Multiple imputations",
author = "Giorgos Bakoyannis and Fotios Siannis and Giota Touloumi",
year = "2010",
month = "12",
day = "30",
doi = "10.1002/sim.4133",
language = "English (US)",
volume = "29",
pages = "3172--3185",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "30",

}

TY - JOUR

T1 - Modelling competing risks data with missing cause of failure

AU - Bakoyannis, Giorgos

AU - Siannis, Fotios

AU - Touloumi, Giota

PY - 2010/12/30

Y1 - 2010/12/30

N2 - When competing risks data arise, information on the actual cause of failure for some subjects might be missing. Therefore, a cause-specific proportional hazards model together with multiple imputation (MI) methods have been used to analyze such data. Modelling the cumulative incidence function is also of interest, and thus we investigate the proportional subdistribution hazards model (Fine and Gray model) together with MI methods as a modelling approach for competing risks data with missing cause of failure. Possible strategies for analyzing such data include the complete case analysis as well as an analysis where the missing causes are classified as an additional failure type. These approaches, however, may produce misleading results in clinical settings. In the present work we investigate the bias of the parameter estimates when fitting the Fine and Gray model in the above modelling approaches. We also apply the MI method and evaluate its comparative performance under various missing data scenarios. Results from simulation experiments showed that there is substantial bias in the estimates when fitting the Fine and Gray model with naive techniques for missing data, under missing at random cause of failure. Compared to those techniques the MI-based method gave estimates with much smaller biases and coverage probabilities of 95 per cent confidence intervals closer to the nominal level. All three methods were also applied on real data modelling time to AIDS or non-AIDS cause of death in HIV-1 infected individuals.

AB - When competing risks data arise, information on the actual cause of failure for some subjects might be missing. Therefore, a cause-specific proportional hazards model together with multiple imputation (MI) methods have been used to analyze such data. Modelling the cumulative incidence function is also of interest, and thus we investigate the proportional subdistribution hazards model (Fine and Gray model) together with MI methods as a modelling approach for competing risks data with missing cause of failure. Possible strategies for analyzing such data include the complete case analysis as well as an analysis where the missing causes are classified as an additional failure type. These approaches, however, may produce misleading results in clinical settings. In the present work we investigate the bias of the parameter estimates when fitting the Fine and Gray model in the above modelling approaches. We also apply the MI method and evaluate its comparative performance under various missing data scenarios. Results from simulation experiments showed that there is substantial bias in the estimates when fitting the Fine and Gray model with naive techniques for missing data, under missing at random cause of failure. Compared to those techniques the MI-based method gave estimates with much smaller biases and coverage probabilities of 95 per cent confidence intervals closer to the nominal level. All three methods were also applied on real data modelling time to AIDS or non-AIDS cause of death in HIV-1 infected individuals.

KW - Competing risks

KW - Cumulative incidence

KW - Fine and Gray model

KW - Missing cause of failure

KW - Multiple imputations

UR - http://www.scopus.com/inward/record.url?scp=78650240622&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78650240622&partnerID=8YFLogxK

U2 - 10.1002/sim.4133

DO - 10.1002/sim.4133

M3 - Article

VL - 29

SP - 3172

EP - 3185

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 30

ER -